Filtering > Wiener filter: predictive deconvolution

The video shows the use of the predictive Wiener filter to eliminate multiples on two synthetic seismic sections of normal incidence.

The first section (without reflectors) is composed of a plane wave with its multiples.
The plane wave can represent the water bottom and its multiples, the multiples of the water bottom. It is possible to add noise to the seismic section by modifying the signal to noise ratio. Even if the water bottom is horizontal, noise varies in distance in this simulation.

The second seismic section (with reflectors) is the previous seismic section to which we added an arrival representing an inclined reflector. The reflected wave is composed of an initial wave and of its multiples created by the multiples on the water bottom. It is possible to select one of the two seismic sections.

For each section, the animation shows on the right:

  • Initial section (Init);
  • Out filtered section (without multiples);
  • Section containing only multiples (multiples);
  • Operator's section (operators).

The animation shows on the left:

  • The average of the autocorrelations of traces of the initial section (Cor-Input) or the average of the autocorrelations of the filtered section (Cor-Output);

The prediction interval is measured on the autocorrelation function of the seismic trace to be deconvolved. The central part of the trace's autocorrelation corresponds here to the autocorrelation of the emitted signal. The autocorrelation of the emitted signal is repeated in time and softened in amplitude. The time difference between the autocorrelation central peak of the trace and the first extremum of opposed sign gives the prediction interval. This type of filter does not require a very precise measurement of the prediction interval.